为保障海上交通安全，本文提出复杂海况下海上船舶运动目标实时检测方法。首先，设计自适应特征强化模块，通过通道强化和空间强化获取船舶目标的特征图，并利用超像素显著性加权平均计算的方式，利用局部区域的颜色、纹理一致性对背景像素进行动态聚类与权重分配，判断目标-背景的类内差异性与类间相似性，初步判定背景区域。结合图像分割能量函数和自适应参数，抑制背景显著性，削弱了海洋动态背景对目标检测的干扰，生成抑制背景后的前景图像；然后，利用颜色共生矩阵矢量和梯度矢量分别描述前景目标的动态特征和静态特征，并引入前向序列选择和后向序列删除的双向优化策略，剔除最劣特征，结合统计度量函数对特征子集进行动态剪枝，筛选出具有最大类间区分度的前景主特征；最后，分别采用快速/慢速运动目标检测模块检测运动目标，融合2个模块的检测结果得到最终检测框，并引入极大值抑制算法去除重复目标框，得到船舶运动目标的最终检测结果。实验结果表明，应用所提方法进行船舶运动目标检测，其交并比阈值高于0.5，检测精度得到提升，论证了设计方法调度的有效性。
In order to ensure the safety of maritime traffic, this study proposes a real-time detection method of marine ship moving target under complex sea conditions. Firstly, the adaptive feature enhancement module is designed to obtain the feature map of the ship target through channel enhancement and spatial enhancement, and the background pixels are dynamically clustered and weighted by using the color and texture consistency of the local region using the method of superpixel saliency weighted average calculation, so as to judge the intra class difference and inter class similarity of the target background, and preliminarily determine the background region. Combining the energy function of image segmentation and adaptive parameters, the background saliency is suppressed, and the interference of the dynamic ocean background on target detection is weakened, and the foreground image after background suppression is generated. Then, the color co-occurrence matrix vector and gradient vector are used to describe the dynamic and static features of the foreground target respectively, and the bidirectional optimization strategy of forward sequence selection and backward sequence deletion is introduced to eliminate the worst feature, and the feature subset is dynamically pruned combined with the statistical measurement function to select the foreground main feature with the maximum inter class discrimination; Finally, the fast/slow moving target detection module is used to detect the moving target, and the final detection frame is obtained by fusing the detection results of the two modules. The maximum suppression algorithm is introduced to remove the duplicate target frame, and the final detection result of the ship moving target is obtained. Experimental results show that applying the proposed method for ship moving target detection achieves an Intersection over Union (IoU) threshold higher than 0.5, improving detection accuracy and demonstrating the effectiveness of the design method scheduling.
2026,48(4): 196-200 收稿日期：2025-7-18
DOI：10.3404/j.issn.1672-7649.2026.04.030
分类号：U672
基金项目：威海市海洋大数据智能应用工程技术研究中心科研开放专项资金项目（HYSJ20240003）
作者简介：张训源(1989-),男,硕士,讲师,研究方向为图像处理、自动控制
参考文献：
[1] 袁学飞, 汪硕晶, 孙庆兵, 等. 基于改进YOLOv7识别算法的船舶检测识别方法研究[J]. 起重运输机械, 2025(3): 52-59.
YUAN X F, WANG S J, SUN Q B, et al. Research on ship detection and recognition method based on improved YOLOv7 algorithm[J]. Hoisting and Transportation Machinery, 2025(3): 52-59.
[2] 黎煜培, 王忠华. 基于局部全局自注意与空间通道稀疏增强的红外船舶目标检测算法研究[J]. 红外与激光工程, 2025, 54(3): 348-360.
LI Y P, WANG Z H. Research on infrared ship target detection algorithm based on local-global self-attention and spatial-channel sparsity enhancement[J]. Infrared and Laser Engineering, 2025, 54(3): 348-360.
[3] 陈燕奎, 龙超活, 何骏杰, 等. Ship-YOLOv8: 一种轻量级高分辨遥感图像船舶细粒度检测算法[J]. 现代信息科技, 2024 (22): 25-30.
CHEN Y R, LONG C H, HE J J, et al. Ship-YOLOv8: a lightweight high-resolution remote sensing image ship fine-grained detection algorithm[J]. Modern Information Technology, 2024, 8(22): 25-30.
[4] 范云生, 张凯, 牛龙辉, 等. 海上无人机对运动船舶的长期检测跟踪算法[J]. 仪器仪表学报, 2024, 45(3): 326-335.
FAN Y S, ZHANG K, NIU L H, et al. Long-term detection and tracking algorithm for moving vessels using maritime UAVs[J]. Journal of Instrument and Measurement, 2024, 45(3): 326-335.
[5] 杨镇宇, 石刘. 基于改进Cascade R-CNN算法的船舶目标检测方法[J]. 舰船科学技术, 2024, 46(6): 144-149.
YANG Z Y, SHI L. A ship target detection method based on the improved cascade R-CNN algorithm[J]. Ship Science and Technology, 2024, 46(6): 144-149.
[6] 丘锐聪, 周海峰, 陈颖, 等. 融合可切换空洞卷积的上下文信息增强船舶目标检测算法[J]. 大连海事大学学报, 2023, 49(4): 116-125.
QIU R C, ZHOU H F, CHEN Y, et al. A context information enhancement algorithm for ship object detection with switchable dilated convolutions[J]. Journal of Dalian Maritime University, 2023, 49(4): 116-125.
[7] 姜杰, 张立民, 刘凯, 等. 基于改进PP-YOLOE和ByteTrack算法的红外船舶目标检测跟踪方法[J]. 兵器装备工程学报, 2024, 45(11): 291-297.
JIANG J, ZHANG L M, LIU K, et al. Infrared ship target detection and tracking method based on improved PP-YOLOE and ByteTrack algorithm[J]. Journal of Ordnance Engineering, 2024, 45(11): 291-297.
[8] 余伟, 尤红建, 胡玉新, 等. 基于多尺度双邻域显著性的高分四号遥感图像运动船舶检测方法[J]. 电子与信息学报, 2023, 45(1): 282-290.
YU W, YOU H J, HU Y X, et al. A high-resolution GaoFen-4 remote sensing image motion ship detection method based on multi-scale dual neighborhood significance[J]. Journal of Electronics and Information Technology, 2023, 45(1): 282-290.
